FtsH-Mediated Coordination of Lipopolysaccharide Biosynthesis in Escherichia coli Correlates with the Growth Rate and the Alarmone (p)ppGpp

Abstract

The outer membrane is the first line of defense for Gram-negative bacteria and serves as a major barrier for antibiotics and other harmful substances. The biosynthesis of lipopolysaccharides (LPS), the essential component of the outer membrane, must be tightly controlled as both too much and too little LPS are toxic. In Escherichia coli, the cellular level of the key enzyme LpxC, which catalyzes the first committed step in LPS biosynthesis, is adjusted by proteolysis carried out by the essential and membrane-bound protease FtsH. Here, we demonstrate that LpxC is degraded in a growth rate-dependent manner with half-lives between 4 min and >2 h. According to the cellular demand for LPS biosynthesis, LpxC is degraded during slow growth but stabilized when cells grow rapidly. Disturbing the balance between LPS and phospholipid biosynthesis in favor of phospholipid production in an E. coli strain encoding a hyperactive FabZ protein abolishes growth rate dependency of LpxC proteolysis. Lack of the alternative sigma factor RpoS or inorganic polyphosphates, which are known to mediate growth rate-dependent gene regulation in E. coli, did not affect proteolysis of LpxC. In contrast, absence of RelA and SpoT, which synthesize the alarmone (p)ppGpp, deregulated LpxC degradation resulting in rapid proteolysis in fast-growing cells and stabilization during slow growth. Our data provide new insights into the essential control of LPS biosynthesis in E. coli.

INTRODUCTION

The outer membrane of Gram-negative bacteria is an asymmetric bilayer, containing phospholipids (PL) in the inner leaflet but PL and lipopolysaccharides (LPS) in the outer leaflet. The LPS molecules act as permeability barrier and confer protection against membrane-disruptive agents (1). The functionality of this protective barrier requires a proper equilibrium between PL and LPS, which is achieved by regulation of the corresponding biosynthesis pathways.

One essential protein in the de novo biosynthesis of fatty acids is the R-3-hydroxyacyl-ACP dehydrase FabZ, which provides precursors for PL assembly (2). FabZ shares its substrate with LpxA and LpxD, which are both involved in the biosynthesis of lipid A. Lipid A is also known as endotoxin and forms the crucial lipid anchor of LPS molecules in the outer membrane (3–5) (see Fig. 2A). The competition of FabZ with LpxA and LpxD for R-3-hydroxyacyl-ACP as the shared precursor is discussed as an important link between PL and LPS biogenesis, coordinating the proper balance of both membrane components. However, it is conceivable that further mechanisms are involved in this critical process. Obviously, not only the relative ratio between PL and LPS but also the quantity of both molecules needs to meet the cellular requirements during different generation times because excessive production of LPS is toxic (6–8). The regulation of LPS synthesis in Escherichia coli and related enterobacteria depends on proteolysis of two essential enzymes, namely, LpxC and KdtA, by the membrane-bound AAA (ATPases associated with various cellular activities) protease FtsH (8, 9). LpxC is a deacetylase, which catalyzes the committed step in biosynthesis of lipid A (10) and KdtA attaches two 3-deoxy-d-manno-octulosonate (Kdo) residues to lipid IV_A to form Kdo₂-lipid A, the minimal structure of LPS that is essential for survival (11). The C terminus of LpxC has been shown to contain a nonpolar, sequence- and length-specific signal for degradation by the FtsH protease (6, 7). Besides its essential role in controlling LPS biosynthesis, FtsH is involved in the quality control of membrane proteins and the regulation of various cellular mechanisms, in particular under stress conditions (12–15). Although about 15 different FtsH substrates are known, there are many open questions concerning their precise degradation mechanism. Usually, proteolysis of key enzymes or regulatory proteins is not constitutive but tightly controlled to allow adaptation to changing growth conditions (16–20). Assuming that the stability of LpxC might be coordinated with the cellular demand for LPS under different growth conditions, we set out to monitor LpxC stability at different generation times. Here, we report that LpxC is degraded rapidly during slow growth, presumably to avoid toxic overproduction of LPS, but is highly stable under optimal growth conditions. We show that LpxC stability and hence LPS production is coupled to the PL biosynthesis machinery. Furthermore, we tested several factors and intracellular signaling molecules for a contribution in LPS biosynthesis and found that the alarmone (p)ppGpp is needed for control of LpxC stability.

Fig 2.

Expression of a hyperactive FabZ protein restored unbalanced LPS synthesis upon LpxC overexpression and resulted in increased stability of LpxC. (A) Scheme of the biosynthesis pathways and localization of phospholipids and lipopolysaccharides. Both pathways share R-3-hydroxyacyl-ACP as precursor used by FabZ or LpxA and LpxD, respectively. (B) In E. coli WT cells, overexpression of LpxC by addition of 0% (white), 0.1% (gray), or 0.5% arabinose (black) resulted in the accumulation of membrane-bound Kdo (middle) compared to WT cells harboring an empty vector (left). In contrast, overexpression of LpxC in a strain encoding for a hyperactive FabZ protein (FabZ*) did not affect the concentration of Kdo in the membranes (right). (C) Microscopic analysis of E. coli WT and FabZ* cells. The addition of 0.5% arabinose for plasmid-derived protein expression did not alter the morphology of WT cells containing a vector control (left) but induced the formation of elongated cells in a strain carrying the expression plasmid for LpxC (middle). The overexpression of LpxC in FabZ* cells does not affect cell morphology (right). Bars, 20 μm. (D). Independent of the generation time, the LpxC protein is stable in FabZ* cells, as monitored by in vivo degradation experiments.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

The bacterial strains used in the present study are listed in Table 1. Strains were cultivated in liquid Luria-Bertani (LB) broth and M9 minimal medium or on respective agar plates. Antibiotics were used in the following concentrations: ampicillin (Amp), 100 μg ml⁻¹; kanamycin (Kan), 50 μg ml⁻¹; chloramphenicol (Cm), 25 μg ml⁻¹; tetracycline, 12.5 μg ml⁻¹; and spectinomycin (Sp), 300 μg ml⁻¹.

Table 1.

Bacterial strains and plasmids used in this study

Strain or plasmid	Characteristics	Source or reference
Escherichia coli strains
W3110 (wild type)	F− IN(rrnD-rrnE)1	76
AR3289 (FabZ*)	F− IN(rrnD-rrnE)1 sfhC21 zad220::Tn10	77
CF5802 (Δppk-ppx)	F− λ− ΔilvG rfb-50 rph-1 Δppk-ppx::Kan	58
CF11608 (SpoT_E319Q)	F− λ− ΔilvG rfb-50 rph-1 spoT(E319Q)a	R. Harinarayanan and M. Cashel
CF1652 (ΔrelA251)	F− λ− ΔilvG rfb-50 rph-1 ΔrelA251::Kan	32
CF1693 (ΔrelA/ΔspoT)	F− λ− ΔilvG rfb-50 rph-1 ΔrelA251::Kan ΔspoT207::Cm	32
RH99 (ΔrpoS359)	F− araD139 Δ(argF-lac)U169 rpsL150 ptsF25 flbB5301 rpsR deoC relA1 Φ(csi-5::lacZ) (λplac Mu55) rpoS359::Tn10	78
Plasmids
pBAD24	Amprb; PBAD; araC	79
pBO110	pBAD24 derivative coding for E. coli LpxC	6

^aMutation results in loss of the (p)ppGpp synthase function of SpoT but does not affect the hydrolase function.

^bAmp^r, ampicillin resistance.

In vivo degradation of LpxC in E. coli.

In vivo degradation experiments were used to measure the stability of plasmid-encoded LpxC proteins in E. coli. Cells carrying the LpxC expression plasmid pBO110 were grown in a water bath shaker (180 rpm) until an optical density at 580 nm (OD₅₈₀) of 0.5 was reached. Temperatures between 23 and 40°C and mixtures of liquid LB and M9 minimal medium were used to adjust different generation times. Expression of LpxC was induced by addition of 0.05% arabinose. After 10 min, translation was blocked by addition of Sp. Samples were taken at different time points to a maximum of 120 min and were frozen in liquid nitrogen.

Protein preparation and LpxC detection.

After thawing on ice, the samples were centrifuged (10 min, 16,000 × g), cell pellets were resuspended in TE buffer (10 mM Tris-HCl [pH 8], 1 mM EDTA) according to their OD₅₈₀ (100 μl for an OD₅₈₀ of 1). Protein sample buffer was added (final concentration: 2% [wt/vol] sodium dodecyl sulfate [SDS]; 0.1% [wt/vol] bromophenol blue; 10% glycerol; and 50 mM Tris-HCl [pH 6.8]), cells were heated (100°C, 10 min) and centrifuged (0.5 min, 16,000 × g). The proteins were separated by SDS-gel electrophoresis and immobilized by Western transfer as described previously (21). Immunodetection was performed using a polyclonal LpxC antibody and a secondary goat anti-rabbit antibody–horseradish peroxidase (HRP) conjugate (Bio-Rad). Chemiluminescence signals were monitored using the Luminata Forte Western HRP substrate (Millipore) and the FluorChem FC2 Imager (Alpha Innotec). The half-lives of LpxC were calculated with the AlphaView software (version 3.0.3.0; Alpha Innotec).

Measurement of Kdo amounts in the membrane.

To monitor cellular LPS production of LpxC-expressing WT and FabZ* cells, relative amounts of membrane-bound Kdo were quantified using a Kdo assay derived from Karkhanis et al. as described previously (7, 22)

Microscopic analysis.

To analyze the effect of LpxC overexpression in WT and FabZ* cells, 2 μl of the induced culture (0.5% arabinose) was fixed in an agarose matrix and inspected with a BX51 microscope (Olympus).

RESULTS

Proteolysis of LpxC is regulated in a growth rate-dependent manner.

As the requirement for membrane biosynthesis increases with the growth rate, we argued that the cellular amount of LpxC might vary to provide appropriate amounts of LPS molecules. To test this assumption, we monitored the degradation of LpxC at various generation times. As endogenous LpxC is barely detectable during normal growth conditions, we measured the half-life of plasmid-encoded LpxC under the control of an arabinose-inducible promoter. E. coli cells harboring pBO110 were grown to exponential phase prior to mild induction of lpxC expression by arabinose. After addition of spectinomycin to block translation, samples were taken at different time points and were analyzed for LpxC abundance by immunodetection using polyclonal antiserum raised against LpxC (6). To adjust generation times, both the temperature and the growth media composition were varied, e.g., cultivation in 50% M9 + 50% LB medium at 39°C resulted in generation times between 32 and 53 min, while growth in M9 medium at 26°C led to a doubling time of up to 333 min. Figure 1A shows representative examples of LpxC half-lives determined in E. coli wild-type (WT) cells at generation times ranging from 32 to 250 min. LpxC was rapidly degraded during slow growth (50 min-6 h doubling time), showing a half-life of about 10 min. This includes generation times of cells grown in LB broth at 30°C, which was previously reported to result in LpxC half-lives of ∼ 10 min (6). LpxC stability increased with increasing growth rate (when the demand for LPS biosynthesis is high), resulting in stabilization over 120 min measured at a generation time of approximately 30 min. Figure 1B summarizes about 30 independent degradation experiments and illustrates the strict correlation between LpxC stability and generation time to adjust LPS biosynthesis in response to the growth rate of the cultures. The growth conditions, resulting generation times and observed LpxC half-lives within these experiments are given in Table 2.

Fig 1.

Growth rate-dependent degradation of LpxC in E. coli WT cells. The stability of plasmid-encoded LpxC (pBO110) was determined in E. coli W3110 cells by in vivo degradation experiments (spectinomycin [Sp] was used to block translation), Western blot analysis, and immunodetection. (A) The decay of LpxC is shown exemplarily for different growth rates. Generation times were adjusted by different growth conditions using varied temperatures and growth media. LpxC half-lives were calculated by densitometric analysis. (B) The summary of about 30 in vivo degradation experiments shows a strong correlation between generation time and LpxC half-life. Open diamonds represent the result from the experiments shown in panel A. The particular growth conditions, generation times, and LpxC half-lives are specified in Table 2.

Table 2.

Growth conditions, the resulting generation times, and LpxC half-lives for degradation experiments in E. coli WT^a

Generation time (min)	Medium	Temp (°C)	LpxC half-life (min)
333	M9	26	9
278	M9	26	9
263	M9	37	8
263	M9	26	9
250	M9	26	9
217	M9	37	11
208	M9	40	9
125	LB	30	4
114	M9	37	11
100	LB	30	6
100	LB	30	8
94	LB	26	10
86	M9	40	8
81	LB	37	17
63	M9 + 10% LB	37	29
55	LB	37	20
55	LB	45	25
54	LB	38	13
53	M9 + 50% LB	39	10
48	LB	39	29
47	M9 + 20% LB	37	21
45	LB	42	46
44	M9 + 50% LB	39	38
41	LB	45	31
38	LB	37	40
37	LB	45	39
35	LB	40	32
33	LB	45	60
32	M9 + 50% LB	39	120
32	LB	40	80
30	LB	40	120

^aConditions used for the representative blots shown in Fig. 1A are highlighted in gray.

Enhanced phospholipid biosynthesis stabilizes LpxC independently of the growth rate.

As an influence of increased FabZ activity on LpxC stability was noted previously (8), we analyzed this relationship in more detail. We made use of an E. coli W3110 strain carrying the sfhC21 mutation, which codes for a hyperactive FabZ protein (strain FabZ*). In an early step of PL biosynthesis, FabZ uses R-3-hydroxyacyl-ACP as a precursor, which is also needed by LpxA and LpxD as acyl-donor in LPS biogenesis (Fig. 2A). Therefore, the hyperactive FabZ protein directs most R-3-hydroxyacyl-ACP to the PL biosynthesis pathway and we set out to analyze the effect on LPS biosynthesis and LpxC stability. Previous work showed that overexpression of active LpxC shifts the equilibrium between LPS and PL causing toxic effects that can be monitored by LPS-bound Kdo accumulation in the membranes and by elongated cell shape (7, 8, 23, 24). Consistent with this, the relative amount of membrane-bound Kdo per cell increased when expression of plasmid-encoded LpxC was induced by increasing arabinose concentrations (Fig. 2B), whereas inducer concentrations up to 0.5% arabinose (black columns) had no effect on cells carrying the empty vector (left columns). In E. coli FabZ* cells, the overproduction of LpxC had no effect on the relative Kdo amount. In addition, the FabZ* strain showed no elongated cell shape during LpxC overexpression (Fig. 2C, right). These results demonstrate that hyperactive FabZ can compensate for elevated LpxC expression and therefore suggests a direct cross talk between PL and LPS synthesis. This link is further supported by the LpxC stability in this strain. LpxC in the FabZ* cells was completely stable and proteolysis was unaffected by long (263 min) or short (41 min) generation times (Fig. 2D).

Loss of RpoS and polyP does not affect growth rate-dependent LpxC degradation.

Different factors are known to contribute to the growth phase-dependent regulation in E. coli. The master regulator of the general stress response is the alternative sigma factor RpoS (25, 26). Therefore, we determined whether a lack of RpoS influences the proteolysis of LpxC by using the E. coli strain ΔrpoS359. In the case of slow growth (doubling time of 185 min) LpxC was degraded with a half-life of 9 min, whereas increased growth rates resulted in stable LpxC protein (Fig. 3A). This pattern in the rpoS deletion strain is comparable to WT cells (Fig. 1B) and demonstrates that RpoS does not affect the growth rate dependency of LpxC proteolysis.

Fig 3.

The absence of the stationary-phase sigma factor RpoS (A) or polyphosphates (B) showed no effect on the growth rate-dependent degradation of LpxC. In vivo degradation experiments of LpxC in a strain lacking RpoS (ΔrpoS359) or lacking polyphosphates by knockout of the polyphosphate kinase (Δppk-ppx) were performed as described above.

Another candidate implicated in both regulated proteolysis and growth rate adaptation is inorganic polyphosphate (polyP). PolyP accumulates mainly when cells grow slowly or enter stationary phase and promotes Lon-dependent turnover of ribosomal proteins for amino acid recycling (27, 28). Both the growth-dependent accumulation and its known function in proteolysis rendered polyP a potential modulator of LpxC degradation. The results of in vivo degradation experiments using E. coli Δppk-ppx cells, which lack both the polyP synthesis and degradation machinery (polyphosphate kinase PPK; exopolyphosphatase PPX) showed no difference in growth-regulated LpxC stability in comparison to E. coli WT cells (compare Fig. 3B and Fig. 1B). Thus, a role of polyP in regulation of LpxC degradation can be excluded.

Decreased amounts of (p)ppGpp deregulate LpxC proteolysis.

Another major player needed for growth adaptation in bacteria is the alarmone (p)ppGpp, which accumulates during slow growth and when cells enter stationary phase (29, 30). It has been reported that lpxC transcription is negatively regulated by (p)ppGpp (31). Here, we analyzed LpxC stability in a set of mutant strains, which synthesize either lower amounts of (p)ppGpp or are void of this signal molecule. Specifically, these strains are transposon mutants defective for the (p)ppGpp synthases RelA, a SpoT mutant that abolishes (p)ppGpp synthase activity but leaves intact (p)ppGpp hydrolase function (point mutation SpoT_E319Q) and a ΔrelA/ΔspoT double mutant.

Interestingly, LpxC degradation in all three mutant strains differed substantially from the pattern in the wild-type background. In E. coli SpoT_E319Q cells, LpxC was degraded with a half-life between 6 and 39 min under all tested generation times (Fig. 4A). Even during fastest growth, the half-life of LpxC did not exceed 30 min. In other words, WT-like stabilization of the protein did not occur and elimination of the synthase activity of SpoT influenced LpxC stability during fast growth.

Fig 4.

Lack of SpoT or RelA affects the growth-dependent degradation of LpxC. (A) In vivo degradation experiments in E. coli cells lacking the synthase function of SpoT (SpoT_E319Q). (B) LpxC was stabilized at very high as well as very low generation times in an E. coli strain lacking the RelA protein (ΔrelA251). For comparison, the degradation pattern of WT cells is indicated by a gray line.

A different effect on LpxC stability was observed in the RelA mutant (Fig. 4B). Although a slight stabilization of LpxC was detectable at short generation times (33 min, half-life of 68 min), LpxC was not rapidly degraded when cells grew slowly. At doubling times of 300 to 400 min, LpxC was much more stable in the mutant than in the WT (half-lives up to 80 min). Thus, LpxC was stabilized both during slow growth and in very fast-growing cells in the RelA-deficient strain.

Finally, we examined degradation of LpxC in a ΔrelA/ΔspoT double mutant, which is unable to synthesize (p)ppGpp (32). As shown in Fig. 5, LpxC was stabilized in slow-growing cells. The maximal LpxC half-life of 94 min was determined for cells growing with a generation time of nearly 390 min. In contrast, very fast-growing cultures (doubling time of 32 min) showed an LpxC half-life of 12 min. The correlation between LpxC stability and generation time as shown for 31 independent degradation experiments in Fig. 5B is reversed in comparison to the growth dependency of LpxC proteolysis in E. coli WT cells. Therefore, either the (p)ppGpp synthases or the cellular concentration of (p)ppGpp itself are crucial for growth-dependent proteolysis of LpxC amounts in E. coli.

Fig 5.

In vivo degradation experiments with an E. coli ΔrelA/ΔspoT double mutant. (A) Densitometric analyses of the presented western transfers show a stable LpxC protein for long generation times. The LpxC stability decreases by shortening the generation time. Within the degradation experiment, various generation times were adjusted by different growth conditions and translation was blocked by the addition of spectinomycin (Sp). (B) Summarizing diagram for in vivo degradation experiments performed under different conditions. The correlation of LpxC half-life and the generation time is contrary to the correlation obtained in E. coli WT cells. The open diamonds correspond to the blots shown in panel A. For comparison, the degradation pattern of WT cells is indicated by a gray line.

DISCUSSION

Tight control of the LPS amount in the outer membrane is critically important for functionality of the physical barrier of Gram-negative bacteria. In E. coli cells grown in LB medium at 30°C, LpxC has previously been shown to be degraded rapidly. By using a wide range of generation times from about 20 to 400 min, we demonstrate now that the stability of this essential enzyme strictly correlates with the growth rate. When cells grow slowly and the cellular need for LPS biosynthesis is low, LpxC is degraded efficiently by FtsH. In good agreement with the increased rate of membrane synthesis when cells double rapidly, LpxC was stabilized under these conditions.

Thus far, conditional proteolysis has only been reported for a few substrates. For instance, (i) increasing temperatures trigger the degradation of the homoserine transsuccinylase HTS (MetA), which catalyzes the first step in the de novo methionine biosynthesis and thereby affects the rate of translation (18), (ii) the stability of the DNA-binding replication inhibitor CspD is adjusted according to the growth rate and growth phase of E. coli cells (19), (iii) anti-sigma factor guided degradation controls the amount of the stationary phase sigma factor RpoS (33), and (iv) the formate dehydrogenase subunit FdoH and the uncharacterized YfgM protein are degraded by FtsH in response to various growth conditions (12).

To learn more about the molecular mechanisms underlying the growth-dependent modulation of LpxC stability, we tested the impact of several intracellular factors. It has been reported for a single generation time that LpxC is stabilized in a strain, which encodes a hyperactive FabZ protein (8). We showed that LpxC is stabilized under all tested conditions in this fabZ allele (Fig. 2). Presumably, this stabilization prevents excess consumption of the joint precursor R-3-hydroxyacyl-ACP by the hyperactive FabZ and avoids toxic effects by shifting the equilibrium between PL and LPS biosynthesis toward LPS production. Ogura et al. showed that (i) 2(E)-tetradecenoyl-ACP, the FabZ product and (ii) inhibition of FabI, the subsequent enzyme in fatty acid biosynthesis, stabilize LpxC as well (8). Earlier studies demonstrated that LpxC activity is affected by inhibition of later steps in lipid A biosynthesis (34, 35). Therefore, it is likely that a cross talk at different stages of both the LPS and PL pathways affects regulated proteolysis of LpxC to orchestrate a balanced biosynthesis of both membrane components. Fatty acid precursors and intermediates also play a regulatory role in these processes (36). For example, the activity of acetyl coenzyme A (acetyl-CoA) carboxylase (ACC) is reduced by accumulation of long-chain acyl-ACPs resulting in increased amounts of acetyl-CoA (37). Remarkably, AccA and AccD subunits were found to interact with the ClpS adaptor protein (38), so these proteins might be targets of ClpAP-mediated proteolysis. Long-chain acyl-ACPs accumulate in high amounts by inactivation of the membrane-bound glycerolphosphate acyltransferase PlsB, which catalyzes the first step in PL biosynthesis (39). Interestingly, PlsB is directly inhibited by (p)ppGpp (40) and therefore links PL biosynthesis to growth adaptation. This connection is underlined by protein interactions between ACP and SpoT (41).

Under nutrient starvation the GTP and ATP-derived alarmone (p)ppGpp is synthesized and controls many cellular functions in the context of the “stringent response” (42, 43). E. coli harbors two (p)ppGpp synthases, the monofunctional synthase RelA and the bifunctional SpoT, which is also able to degrade the signal molecule to GTP or GDP and PP_i (43, 44). The ribosome-associated RelA is activated when cells starve for amino acids and uncharged tRNAs accumulate (45). SpoT activity responds to a variety of starvation conditions, including the lack of nitrogen, carbon, fatty acids, phosphate, and iron (46–50). (p)ppGpp was shown to accumulate both in stationary phase and in slow-growth exponential phase (29, 30) to influence a multitude of cellular functions mediating the adaptation to limiting growth conditions. The major target of (p)ppGpp is the RNA polymerase (RNAP). As a consequence, transcription of numerous genes can be up- or downregulated by binding of the alarmone and the ppGpp- and autoregulated RNAP-binding protein DksA (31, 51, 52). The regulated genes include lpxC and kdtA, which both are downregulated during (p)ppGpp accumulation. This reduces the efficiency of LPS biosynthesis already at the transcriptional level. Further targets of (p)ppGpp under starvation conditions are translation, replication, the flagellar machinery or biosynthesis of fatty acids and phospholipids by inhibition of FabA and PlsB (40, 53–57). A major player in the context of growth control is the exopolyphosphatase PPX, which is directly inhibited by (p)ppGpp, resulting in accumulation of inorganic polyphosphates (58). These linear polymers of tens or hundreds of phosphate molecules are essential to trigger the adaptation to stationary phase and stimulate the degradation of ribosomal proteins by the AAA⁺ Lon protease (27, 28). Since (p)ppGpp controls synthesis of anti-adaptor proteins, RpoS stability is directly linked to the (p)ppGpp level in the cell and therefore to the regulation of growth (25, 59–62). Our data exclude an effect of inorganic polyphosphates and RpoS on LpxC stability. In contrast, we found that the stability of LpxC is linked to the presence of RelA and SpoT and therefore to the ability of (p)ppGpp production. The half-life in the single mutant strains (Fig. 4) revealed a different dependency of LpxC stability on growth rate, whereas regulation of LpxC degradation is completely reversed in a ΔrelA/ΔspoT double transposon strain compared to WT cells (Fig. 1 and Fig. 5). A direct connection between (p)ppGpp levels and general proteolysis is known, as global protein degradation increases proportionally with (p)ppGpp expression to provide amino acids for new proteins needed for growth adaptation (63). Recently, a direct link between FtsH and RelA activity was shown for a proteolytic cascade controlling production of a intercellular signaling molecule in Myxococcus xanthus (64). Earlier studies revealed different FtsH amounts in E. coli relA and spoT knock out strains. Slominska et al. found increased amounts of FtsH in a ΔrelA/ΔspoT double mutant as well as in a ΔrelA single mutant compared to WT cells. In addition, the cellular amount of FtsH is correlated to (p)ppGpp level as the highest amount of FtsH is detected in the case of very low and very high (p)ppGpp concentrations. In between those maxima, the concentration of FtsH is decreased (65). This indicates that higher expression levels of ftsH might trigger proteolysis of LpxC under slow-growth conditions. However, differential proteolysis cannot be explained by altered FtsH abundances alone, since FtsH amounts are also higher when LpxC is stabilized. Therefore, the control of LpxC degradation is most likely a multilayered mechanism, which ensures a very sensitive modulation of LpxC levels in the cell. For example, binding of (p)ppGpp could influence the activity of FtsH, as shown for other proteins besides the RNA polymerase as the main target of this molecule (66, 67). However, a typical GTP binding side (68) indicative of (p)ppGpp binding is not present in the structures of LpxC or FtsH. Moreover, isothermal titration calorimetry with FtsH or LpxC did not support a direct interaction with ppGpp (data not shown).

The differential degradation patterns in all three tested (p)ppGpp synthase mutant strains indicate that not only the cellular concentration of the signal molecule but also a regulatory effect of each of both synthesizing enzymes accounts for differential stability of LpxC. Using the bacterial two-hybrid system we screened for the interaction of RelA, SpoT, or ACP with LpxC. ACP was chosen, because this protein is involved in LPS synthesis by providing fatty acids and since it was previously shown to interact with PlsB, LpxD, and SpoT (41, 69, 70). However, no interaction with LpxC was observed (data not shown). Therefore, we conclude that the variations in LpxC stability most likely are not based on a direct interplay with RelA, SpoT, and ACP.

As a novel finding we report that LPS biosynthesis is linked to the cellular growth rate by adjustment of LpxC stability. Future studies will have to elucidate additionally factors involved, such as adaptor proteins that modulate proteolysis of LpxC. Uncoupling of LpxC proteolysis has been discussed previously as a target for antibiotic drug design (71–75). However, we showed recently that regulated proteolysis of LpxC by FtsH is not conserved in all Gram-negative bacteria (24). The LpxC protein of some bacteria is either degraded by a different protease or is stable and probably regulated on the activity level. Since (p)ppGpp is a conserved alarmone in nearly all bacteria, the newly discovered link between LPS biosynthesis and (p)ppGpp promises potential for further research.